About PLXY

PLXY AI Call Center is built and operated by PLXY for production telephony in Cambodia — an AI callbot between the national Telco SIP trunk and a 3CX agent pool, with the studio this documentation describes on top.

Founder

PLXY is designed and engineered by Seng Phirum — founder and lead engineer at PLXY, and the author of KNetraHub, whose documentation style (and in-browser AI assistant) this site follows. Reach out on GitHub or by email.

The shape of the codebase

Five separated projects and one shared library, each self-contained and liftable into its own repository (ADR-0018). The separation is not aesthetic: the gateway half runs on a DMZ host, the studio half scales independently, and the one shared piece — the scenario schema — is shared precisely because two copies of a document contract will eventually disagree (ADR-0022).

How decisions are kept

Every consequential choice is an Architecture Decision Record in docs/adr — forty and counting, from stack pins to slot-filling semantics. When behavior surprises you, the ADR usually explains why it is that way on purpose.

This documentation

• Concept: sidebar manual + one overview page, in the style of KNetraHub's documentation.
• Source: the docs/ directory of the repository — the architecture, setup guide, and ADRs in the sidebar are the same files engineers read in the repo.
• Build: VitePress; ./plxy.sh docs locally, the docs GitHub Action for publishing. The system build (verify, build-docker) never touches it.

Brand

The mark is the Voice Graph: a speech waveform whose peaks are flow nodes joined by a routing thread — the call and the scenario builder in one shape, anchored on dodgerblue #1e90ff.